Shares of AMD climbed 11.7% over four sessions to $245.94 after the company helped ratify new open-standard specifications for connecting AI chips inside data centers — a direct challenge to Nvidia's proprietary networking grip. On April 7, the UALink Consortium announced the ratification of its latest specifications, adding in-network compute, chiplet integration, and manageability features. The market read it as AMD staking a credible claim in AI infrastructure beyond just selling chips.
An Open Standard Backed by Nvidia's Biggest Customers
This isn't a fringe effort. The consortium board includes AMD, Apple, AWS, Google, Intel, Meta, and Microsoft — essentially every major cloud buyer except Nvidia. With over 115 member companies and products expected in 2027 , the coalition signals that hyperscalers want bargaining power against Nvidia's lock-in. For AMD, which led the creation of the UALink Consortium , the payoff is positioning its GPUs as the natural beneficiary of an open ecosystem.
The Revenue Reality: Specs Are Not Silicon
Investors should temper enthusiasm. UALink 1.0 chips won't reach labs until the second half of 2026 and won't appear in products until 2027.
The consortium chair admits versions 1.0 and 2.0 won't fully compete with Nvidia, and only version 3.0 — due next year — will achieve performance parity. Meanwhile, Nvidia's networking revenue surged 142% to $31.4 billion in its latest fiscal year , and nearly 90% of customers buying Nvidia AI systems also purchase its networking products.
AMD's AI Business Is Growing — But From a Distant Second Place
AMD reported record Q4 2025 results: $10.3 billion in quarterly revenue, up 34%, with $5.4 billion in data center revenue alone.
Yet AMD holds just 5–8% of the AI accelerator market versus Nvidia's roughly 80%. The stock trades at ~30x forward earnings, a multiple justified only if 60%+ data center growth materializes.
Nvidia Isn't Standing Still on Openness
Nvidia introduced NVLink Fusion last year, broadening access to its interconnect beyond Nvidia-only GPU deployments.
With gross margins topping 70% , Nvidia has pricing room to defend its turf. The real question: will data center operators actually switch to an unproven open standard when Nvidia's walled garden keeps delivering performance? AMD's stock rally prices in a future where they do. The silicon to prove it hasn't shipped yet.